- Email: [email protected]
Atomic-scale distribution of impurities in CuInSe2-based thin-film solar cells
Ultramicroscopy 111 (2011) 552–556
Contents lists available at ScienceDirect
Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic
Atomic-scale distribution of impurities in CuInSe2-based thin-film solar cells O. Cojocaru-Mire´din a,n, P. Choi a, R. Wuerz b, D. Raabe a a b
Max-Planck-Institut f¨ ur Eisenforschung, Max-Planck-Straße 1, 40237 D¨ usseldorf, Germany Zentrum f¨ ur Sonnenenergie- und Wasserstoff-Forschung Baden-W¨ urttemberg, Industriestraße 6, 70565 Stuttgart, Germany
a r t i c l e i n f o
a b s t r a c t
Available online 11 January 2011
Atom Probe Tomography was employed to investigate the distribution of impurities, in particular sodium and oxygen, in a CuInSe2-based thin-film solar cell. It could be shown that sodium, oxygen, and silicon diffuse from the soda lime glass substrate into the CuInSe2 film and accumulate at the grain boundaries. Highly dilute concentrations of sodium and oxygen were measured in the bulk. Selenium was found to be depleted at the grain boundaries. These observations could be confirmed by complementary energy dispersive X-ray spectroscopy studies. Our results support the model proposed by Kronik et al. (1998) [1], which explains the enhanced photovoltaic efficiency of sodium containing CuInSe2 solar cells by the passivation of selenium vacancies at grain boundaries. & 2011 Elsevier B.V. All rights reserved.
Keywords: Thin-film solar cells Photovoltaic Cu(In,Ga)Se2 CuInSe2 Sodium diffusion Grain boundary segregation Pulsed laser atom probe Transmission electron microscopy
1. Introduction Copper indium diselenide (CIS), a compound semiconductor belonging to the I–III–VI2 chalcopyrite family, is one of the most promising absorber materials for the fabrication of thin-film solar cells. CIS possesses several beneficial material properties such as a direct band gap (Eg ¼1.02 eV at room temperature), high absorption coefficient ( Z105 cm 1), long-term stability, and high radiation hardness [2]. CIS solar cells are commonly deposited on soda lime glass (SLG) substrates, which comprise SiO2, Na2O, K2O, and CaO. Several authors reported diffusion of alkali impurities, in particular sodium, into the CIS film during the processing of these solar cells [3,4]. Interestingly, sodium impurities were found to significantly enhance the photovoltaic conversion efficiency (mainly the open circuit voltage) of CIS solar cells, also known as the ‘‘sodium effect’’ [5,6]. An increase in p-type conductivity [7] and a more pronounced texture (predominance of uniaxial crystallite orientation) [8] were suggested as possible reasons for the enhanced efficiency due to sodium incorporation. However, the ‘‘sodium effect’’ has remained a subject of controversial discussion. To develop a comprehensive theory of how sodium improves the conversion efficiency of CIS solar cells, one needs to study the elemental distribution at the atomic scale. The concentrations in bulk and grain boundaries are of particular interest. Some attempts have been previously made to determine the impurity concentrations in CIS absorber layers. For instance, Niles et al. [9] applied Auger Electron Spectroscopy (AES) to detect sodium and oxygen segregation at the surface and grain boundaries of CIS and
n
Corresponding author. E-mail address: [email protected] (O. Cojocaru-Mire´din).
0304-3991/$ - see front matter & 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.ultramic.2010.12.034
to show that solute concentrations inside the grains are highly dilute. However, the detection limit of AES for both sodium and oxygen was reported to be only about 0.15 at% [9]. Among the nano-analytical techniques, Atom Probe Tomography (APT) [10–13] is one of the most promising tools for studying solar cells, as it allows for spatially resolved chemical analyses with subnanometer resolution [14] and the detection of dilute impurity concentrations as low as a few tens of ppm [15]. First APT results on the elemental distribution in Cu(In,Ga)Se2 layers, which are similar to CIS, were recently published by Cadel et al. [15] and Schlesiger et al. [16]. Both authors reported significant sodium enrichment at the grain boundaries. The current paper elucidates the atomic-scale elemental distribution in an actual CIS solar cell, using pulsed laser APT. The enhanced energy conversion efficiency due to sodium incorporation is discussed on the basis of the obtained APT results.
2. Experimental ¨ SonnenA CIS solar cell was fabricated at the Zentrum fur energie- und Wasserstoff-Forschung in Stuttgart, Germany. Soda lime glass (SLG) of 3 mm in thickness was used as a substrate material. A Mo back contact layer of 500 nm was deposited onto the SLG substrate by DC sputtering. Subsequently, a polycrystalline CuInSe2 film of 2 mm in thickness was grown by co-evaporation of the constituent elements using a high temperature single-stage inline process [17]. The substrate temperature during film growth was kept at 600 1C for about 30 min. A CdS buffer layer was deposited by means of chemical bath deposition at 65 1C followed by RF sputtering of an intrinsic ZnO layer, DC sputtering of a ZnO:Al front contact layer, and electron beam evaporation of Ni/Al contact grids.
O. Cojocaru-Mire´din et al. / Ultramicroscopy 111 (2011) 552–556
X-ray fluorescence spectrometry (XRF) measurements were performed to determine the overall composition of the CIS layer, using an EAGLE XXL system (0.1 mbar, Si(Li) detector and 50 kV Rh X-ray source). Using the Cu, In, and Se K lines, the accuracy of the calculated concentrations is better than 0.5 at%. Structural and compositional analyses of CIS films were done by scanning transmission electron microscopy (STEM) and atom probe tomography (APT), respectively. For the STEM investigations, cross-sectional samples were prepared using a dual-beam focused-ion-beam (FIB) (FEI Helios Nanolab 600) and the lift-out technique [18]. STEM was done at a JEOL 2200 FS microscope operated at 200 kV. APT samples were also prepared by means of FIB according to the procedure described in [19,20]. In order to minimize beam damage, a low energy (5 keV) Ga beam was used at the final ion-milling stage. All APT experiments were carried out applying laser pulses of 532 nm wavelength, 12 ps pulse length, and an energy of 0.1 nJ per pulse at a repetition rate of 100 kHz. The specimen base temperature was about 60 K.
3. Results and discussion 3.1. Microstructure of the CIS layer Fig. 1 shows a cross-sectional STEM bright-field image of the studied CIS solar cell, where the stacking sequence of the constituent layers is ZnO:Al/ZnO/CdS/CuInSe2/Mo/SLG. The STEM image reveals
Fig. 1. STEM bright-field image of a CuInSe2-based thin-film solar cell in a crosssectional view.
553
a polycrystalline CIS layer, which is not smooth but exhibits substantial roughness. Top-view scanning electron microscopy images (not shown here) revealed faceted grains of large size (between 0.7 and 1.25 mm) and are in agreement with STEM observations. Besides the large grains in the top CIS region, small grains of 0.13 to 0.36 mm in size can be recognized at the bottom of the CIS layer (see Fig. 1). Such a bimodal grain structure is typical of CIS films grown by elemental co-evaporation in the presence of sodium [21] and can be discussed in terms of the van der Drift evolutionary selection mechanism [22]. In this mechanism, the nucleation stage is characterized by randomly oriented nuclei that initially grow independently at uniform rate. In the following stage, the crystals with the direction of fastest growth perpendicular to the substrate are favored over other crystals and will prevail. For a tetragonal crystal, the directions of fastest growth are /0 0 1S, /1 0 0S, and /1 1 0S [22]. Another interesting feature is the presence of a few voids (about 10 to 100 nm in size) in the CIS layer close to the CIS/Mo interface (see Fig. 1). These voids probably stem from the deposition process and are not formed during TEM sample preparation, as the sample was finally ion-milled at only 5 kV. Such nano-voids in the CIS layer were also observed by Lei et al. [23] and for CIGS layers by Rudmann [24].
3.2. Composition of CIS layer (grain and grain boundary) A typical mass spectrum obtained from an APT analysis of a CIS layer is given in Fig. 2. While copper and indium were detected as both single and double charged ions, selenium ions were only detected in a single charged state. Complex ions, namely CuCO + , CuSe + , Cu2Se + , CuSe2+ , Se2+ , and Se3+ were detected as well. Sodium was detected as Na + . The background noise level in the mass range close to the Na + peak is extremely low (E40 ppm/a.m.u.) so that even very low dilute sodium concentrations can be detected. Besides sodium, oxygen and silicon impurities were detected as O + , O2+ , Si2 + , and Si + . A small amount of Ga + was also detected. Fig. 3 shows three-dimensional maps of all the detected elements as well as sodium and oxygen only. The analyzed volume is about 120 120 390 nm3 in size and represents the interior of a CIS grain. The measured bulk composition is 23.470.06 at% Cu, 29.070.07 at% In, and 47.470.08 at% Se, which is in good agreement with the integral XRF measurement (compare APT-1 and
Fig. 2. Mass spectrum of the complete measurement from Fig. 3 with details showing the minority elements (Na and O) in the low mass region.
O. Cojocaru-Mire´din et al. / Ultramicroscopy 111 (2011) 552–556
554
XRF in Table 1). In the following, APT-1 and APT-2 denotes the APT measurements presented in this work. While the sodium concentration detected by APT for the CIS bulk is only about 0.002 at% (20 ppm), the oxygen concentration is substantially higher and is about 0.07 at% (700 ppm). The sodium concentration measured in this work is significantly lower than the value reported by Cadel et al. [15] for a Cu(In,Ga)Se2 (CIGS) film (85 ppm). Furthermore, Cadel et al. [15] did not report any oxygen inside CIGS. The difference in the results between these two works may well be ascribed to different processing conditions as well as to different compositions of the absorber layers (CIS and CIGS). In Fig. 3 neither sodium nor oxygen clusters can be distinguished. If such clusters exist they are so tiny that they cannot be clearly recognized in the elemental maps. Thus, to verify or refute a possible clustering trend or short range ordering of sodium and/ or oxygen atoms, the cluster identification (maximum separation method [25]) and the first nearest neighbor (abbreviated as 1NN) distance distribution method were applied. Neither sodium nor oxygen clusters were identified by the cluster identification, even by systematically varying Nmin (minimum number of atoms per cluster) and dmax (maximum distance between sodium and/or oxygen atoms) values. The measured 1NN distance distribution for the oxygen atoms, which is compared to a randomized oxygen distribution in Fig. 4, should therefore be interpreted as a random oxygen distribution. The same trend is observed for Na (not shown here). Fig. 5a shows another APT data set of the CIS layer. In these elemental maps, we clearly recognize sodium, oxygen, and silicon enriched zones, and also a strong correlation between the distributions of these elements. Rotation and careful inspection of these enrichment zones reveal that they have a two-
Fig. 3. APT elemental maps of copper (blue), indium (pink), selenium (red), gallium (yellow), sodium (maroon) and oxygen (sky blue). The volume size shown in this figure is 120 120 390 nm3. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
dimensional shape. However, these enrichments are not continuous but are divided into four plate-shaped parts, marked in Fig. 5a by a dashed-line (for more details see Ref. [26]). Previous AES studies revealed a correlation between sodium and oxygen enrichments (in agreement with Fig. 5a) at the CIS grain boundaries [9]. Taking these previous results and the two-dimensional shape of the enrichments into account, we may conclude that the solute enrichments seen in Fig. 5a are located at or close to a grain boundary. The measured bulk composition (excluding the grain boundary zone) is 21.070.06 at% Cu, 28.8 70.07 at% In, and 50.0 70.08 at% Se, which is in better agreement with the integral XRF measurement (compare APT-2 and XRF in Table 1) than the APT-1 composition. Our APT measurements indicate compositional fluctuations within the CIS film. This result is in agreement with studies done by Eich et al. [29] on lateral inhomogeneities of a Cu(In,Ga)Se2 absorber film. Slight variations in sodium and oxygen concentrations in the bulk compared to APT-1 can be measured as well (2675 ppm of sodium and 600 710 ppm of oxygen, respectively). Although sodium and oxygen atoms are strongly enriched at the grain boundaries, their concentration rapidly drops to the bulk values, when moving away from the segregation zone. In the bulk, sodium and oxygen atoms are homogeneously distributed over several hundreds of nanometers, indicating that long range diffusion has taken place. These observations suggest that the measured bulk concentration values of sodium and oxygen correspond to their solubility limit in CIS. In all APT measurements small amounts of gallium were detected ( 0.1 at%). Gallium atoms stem from the deposition process, as confirmed by secondary ion mass spectrometry (SIMS) ( 0.04 at%). A possible explanation for the relatively high gallium content measured by APT is the implantation of gallium
Fig. 4. Distribution of the 1NN distances for the oxygen atoms, represented in Fig. 3. Comparison with the randomised oxygen distribution. The bin size used for this distribution is 0.2 nm.
Table 1 Chemical composition of single CIS grains measured by APT and of the whole CIS layer (integral) measured by XRF; APT-1: composition of grain shown in Fig. 3 and APT-2: composition of grain shown in Fig. 5 without GB area. C(at%)
CCu
CIn
CSe
CNa
CO
CGa
CSi
APT-1 APT-2 XRF
23.4 7 0.06 21.0 7 0.06 21.3
29.07 0.07 28.8 70.07 27.7
47.4 7 0.08 50.0 70.08 50.8
0.0027 0.0005 0.00267 0.0005 –
0.077 0.001 0.067 0.001 –
0.12 0.11 0.02
– – –
O. Cojocaru-Mire´din et al. / Ultramicroscopy 111 (2011) 552–556
555
of a grain boundary ( 0.5 nm), which can be partly ascribed to a local magnification effect [27]. This reconstruction artefact results in a magnified width and lower atomic density of the oxygen rich grain boundary region, which has a higher evaporation field than the CIS matrix. Letellier [28] observed for nickel-based superalloys that the influence of the local magnification effect increases continuously as the grain boundary approaches an orientation parallel to the sample tip axis. Thus, the enlarged width of the oxygen rich grain boundary zone in Fig. 5a and b is not surprising. The Gibbsian interfacial excess of solutes G (number of segregated atoms per unit interfacial area) was determined from the proxigrams following the procedure described in [29,31]. We obtain GO ¼4.7 1019 atoms/m2, GNa ¼1.4 1018 atoms/m2, and GSi ¼2.8 1018 atoms/m2. These values correspond to nearly 10, 0.25, and 0.5 monolayers of oxygen, sodium, and silicon, respectively, with respect to the (1 0 0) CIS plane, where 10 monolayers of oxygen correspond to 5 nm. Hence, an oxygen containing compound phase may have been formed at the grain boundary. It is not possible to determine the exact composition of the oxygen rich zones because of the strong local magnification effect. However, a strong depletion of selenium can be clearly detected in conjunction with the oxygen enrichment, strongly suggesting that a large number of vacant selenium sites at the grain boundaries are occupied by oxygen atoms. It should be mentioned that similar trends could be measured for several grain boundaries using EDX (not shown here). APT analyses of other grain boundaries confirmed sodium and oxygen segregation. On the other hand, compositional differences at the grain boundaries could be detected as well. For instance, the slight copper enrichment and indium depletion seen in Fig. 5b is not generally observed. The difference in the measured compositions of various grain boundaries is not surprising in view of the variation in solute contents as a function of grain boundary misorientation [32].
3.3. Correlation between elemental distribution and enhanced photovoltaic efficiency
Fig. 5. (a) APT elemental maps of copper (blue), indium (pink), selenium (red), gallium (yellow), sodium (maroon), oxygen (sky blue) and silicon (black). The volume size shown is 72 72 260 nm3. (b) Proxigram of species in the sample with respect to the O 7% isoconcentration surface; the bin size is 0.2 nm. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
during focused-ion-beam milling. However, we note that the implantation zone in CIS specimens is restricted to a few nanometers below the surface if 5 keV ions are used in the final ionmilling step. Fig. 5b shows a proxigram (for more details see Ref. [30]) of the elements detected in the sample with respect to the 7 at% oxygen isoconcentration surface. The proxygram clearly reveals sodium, oxygen, and silicon accumulations at a CIS grain boundary, where the oxygen concentration is substantially higher ( 18 at%) than the sodium ( 0.6 at%) and silicon (1.4 at%) concentration ( 1 at%). These solute enrichments are most likely due to out-diffusion from the soda lime glass substrate during the deposition process (at 600 1C for 30 min), as they were not found in APT measurements of CIGS layers grown on steel substrates (not shown here, to be published). The width of the detected oxygen enrichment is about 7 to 8 nm and therefore much larger than the typical structural width
Our APT data clearly reveal that sodium, silicon, and oxygen atoms diffuse from the SLG substrate into the CIS film during film growth. The enrichment of sodium and oxygen at the grain boundary indicates that diffusion into the CIS film occurs mainly along the grain boundaries, where defect density is higher than in the bulk. Highly dilute concentrations of sodium ( 20 ppm) and oxygen ( 700 ppm) are detected inside the CIS grains. These solutes could passivate detrimental lattice defects (e.g., donor like defects like selenium vacancies, indium atoms at copper sites, etc.) and contribute to the enhancement of the solar cell efficiency. However, the dilute concentrations of the solutes in the grains indicate the ‘‘sodium effect’’ cannot be purely regarded as a bulk effect. As presented in Fig. 5 grain boundary enrichments of oxygen and sodium are accompanied by a strong depletion of Se. These observations are in agreement with the model proposed by Kronik et al. [1] explaining the ‘‘sodium effect’’ on the basis of the well-known catalytic effect of alkali metals on surface oxidation of semiconductors [33]. O2 can be formed from physisorbed O2 by alkali-metal induced polarization of the O–O bond and lowering of the surface work function. According to Ref. [1], the formation of grain boundaries in CIS gives rise to a large number of Se vacancies, which introduce intragap donor levels. These donors lower the effective p-type conductivity. Neutralizing the selenium vacancies by chemisorbed O2 according to the reaction + [In0–V+Se ] +O2 -[In0–O0Se], the photovoltaic efficiency can be increased. Once formed, the In–O bonds are stable and can only
556
O. Cojocaru-Mire´din et al. / Ultramicroscopy 111 (2011) 552–556
be broken by significant and deliberate reduction. Furthermore, this defect chemical model and our APT results suggest that not only grain boundary passivation is enhanced by the presence and catalytic effects of sodium but also sodium diffusion can be assisted by oxygen.
4. Conclusions In this work, Atom Probe Tomography was employed to investigate the elemental distribution in the CuInSe2 absorber of a thin-film solar cell. Sodium and oxygen atoms were found to be distributed homogeneously in the CuInSe2 grains. The measured concentrations of sodium ( 0.002 at%) and oxygen ( 0.07 at%) may well correspond to the solubility limit of sodium and oxygen in CuInSe2 at 600 1C. Furthermore, a correlated enrichment of sodium, oxygen, and silicon atoms was detected, presumably at a grain boundary. These enrichments are accompanied by selenium depletion, which is in good agreement with the defect chemical model proposed by Kronik et al. [1]. The ‘‘sodium effect’’ in our CIS-based solar cell could be explained by the passivation of selenium vacancies (detrimental donor defects) by oxygen.
Acknowledgements The authors thank Wolfram Witte for the CIS-based solar cell preparation and also Aleksander Kostka and Axel Eicke for their help in this work. References [1] L. Kronik, D. Cahen, H.W. Schock, Adv. Mater. 10 (1998) 31. [2] Hegedus, Antonio Luque, Steven, Handbook of Photovoltaic Science and Engineering, England: s.n., 2003. ¨ ˚ [3] J. Hedstrom, H. Ohlsen, M. Bodegard, A. Kylner, L. Stolt, D. Hariskos, M. Ruckh, H.W. Schock, Conf. Rec. 23rd IEEE Photovolt. Spec. Conf. IEEE, Piscataway, N.J., 1993, p. 364. ¨ [4] L. Stolt, J. Hedstrom, J. Kessler, M. Ruckh, K.-O. Velthaus, H.-W. Schock, Appl. Phys. Lett. 62 (1993) 597. ˚ [5] K. Granath, M. Bodegard, L. Stolt, Sol. Energy Mater. Sol. Cells 60 (2000) 279–293. ˚ [6] M. Bodegard, K. Granath, L. Stolt, Thin Solid Films 361 (2000) 9–16.
[7] J. Holz, F. Karg, H. von Philipsborn, in: Proceedings of 12th European Photovoltaic Solar Energy Conference, Amsterdam: s.n., 1994, p. 1592. [8] B.J. Stanbery, Critical Rev. Sol. Mater. Sci. 27 (2) (2002) 73–117. [9] D.W. Niles, M. Al-Jassim, K. Ramanathan, J. Vac. Sci. Technol. A 17 (1) (1999) 291. [10] A. Cerezo, T.J. Godfrey, S.J. Sijbrandij, G.D.W. Smith, P.J. Warren, Rev. Sci. Instrum. 69 (1998) 49. [11] M.K. Miller, T.F. Kelly, Rev. Sci. Instrum. 78 (2007) 031101. [12] B. Deconihout, F. Vurpillot, B. Gault, G. Da Costa, M. Bouet, A. Bostel, A. Hideur, G. Martel, M. Brunel, D. Blavette, Surf. Interf. Anal. 39 (2007) 278–282. [13] B. Gault, F. Vurpillot, M. Gilbert, A. Vella, A. Menand, D. Blavette, B. Deconihout, Rev. Sci. Instrum. 77 (2006) 043705. [14] B. Gault, M.P. Moody, F. De Geuser, A. La Fontaine, L.T. Stephenson, D. Haley, S.P. Ringer, Microsc. Microanal. 16 (2010) 99–110. [15] E. Cadel, N. Barreau, J. Kessler, P. Pareige, Acta Mater. 58 (2010) 2634–2637. ¨ [16] R. Schlesiger, C. Oberdorfer, R. Wurz, G. Greiwe, P. Stender, M. Artmeier, P. Pelka, F. Spaleck, G. Schmitz, Rev. Sci. Instrum. 81 (2010) 043703. [17] M. Powalla, G. Voorwinden, B. Dimmler. in: Proceedings of the 14th European Photovoltaic Solar Energy Conference, Barcelona, Spain: s.n., 1997, p. 1270. [18] L.A. Giannuzzi, J.L. Drown, S.R. Brown, R.B. Irwin, F.A. Stevie, Microsc. Res. Tech. 41 (4) (1998) 285–290. [19] M.K. Miller, K.F. Russell, K. Thompson, R. Alvis, D.J. Larson, Microsc. Microanal. 13 (2007) 428–436. [20] K. Thompson, D. Lawrence, D.J. Larson, J.D. Olson, T.F. Kelly, B. Gorman, Ultramicroscopy 107 (2007) 131–139. ˚ [21] M. Bodegard, O. Lundberg, J. Lu, L. Stolt, Thin Solid Films 431–432 (2003) 46–52. [22] A. van der Drift, Philips Res. Rep. 22 (1967) 267–288. [23] C. Lei, A. Rockett, I.M. Robertson, J. Appl. Phys. 100 (2006) 073518. [24] D. Rudmann, Effects of sodium on growth and properties of Cu(In, Ga)Se2 ¨ thin films and solar cells, Ph.D Thesis, Eidgenossische Technische Hochschule ¨ ETH Zurich, Nr. 15576, 2004. [25] M.K. Miller, Atom Probe Tomography, [Ed.] Academic/Plenum Press, New York: s.n., 2000. [26] O. Cojocaru-Miredin, P. Choi, R. Wuerz, D. Raabe, in: S. Vlase, and A. Chiru D. Bigoni, [Eds.], Proceedings of Third International Conference on Advanced Composite Materials in Civil Engineering, Brasov/Romania: Transilvania University, 2010,’’ vol. 1, p. 45. [27] F. Vurpillot, A. Cerezo, D. Blavette, D.J. Larson, Microsc. Microanal. 10 (3) (2004) 384–390. [28] L. Letellier, Etude des joints de grains et interphases dans les superalliages Astroloy par microscopie electronique et tomographie atomique, University of Rouen, Rouen, France, 1994. [29] D. Eich, U. Herber, U. Groh, U. Stahl, C. Heske, M. Marsi, M. Kiskinova, W. Riedl, R. Fink, E. Umbach, Thin Solid Films 361–362 (2000) 258–262. [30] O.C. Hellman, J.A. Vandenbroucke, J. Rusing, D. Isheim, D.N. Seidman, Microsc. Microanal. 6 (2000) 437–444. [31] O.C. Hellman, D.N. Seidman, Mater. Sci. Eng. A 237 (2002) 24–28. [32] A. Biscondi, M. Fraczkiewicz, J. Physique 46 (1985) 497. [33] H.I. Starnberg, P. Soukisasian, Z. Hurych, Phys. Rev. B 39 (1989) 12775.
Contents lists available at ScienceDirect
Ultramicroscopy journal homepage: www.elsevier.com/locate/ultramic
Atomic-scale distribution of impurities in CuInSe2-based thin-film solar cells O. Cojocaru-Mire´din a,n, P. Choi a, R. Wuerz b, D. Raabe a a b
Max-Planck-Institut f¨ ur Eisenforschung, Max-Planck-Straße 1, 40237 D¨ usseldorf, Germany Zentrum f¨ ur Sonnenenergie- und Wasserstoff-Forschung Baden-W¨ urttemberg, Industriestraße 6, 70565 Stuttgart, Germany
a r t i c l e i n f o
a b s t r a c t
Available online 11 January 2011
Atom Probe Tomography was employed to investigate the distribution of impurities, in particular sodium and oxygen, in a CuInSe2-based thin-film solar cell. It could be shown that sodium, oxygen, and silicon diffuse from the soda lime glass substrate into the CuInSe2 film and accumulate at the grain boundaries. Highly dilute concentrations of sodium and oxygen were measured in the bulk. Selenium was found to be depleted at the grain boundaries. These observations could be confirmed by complementary energy dispersive X-ray spectroscopy studies. Our results support the model proposed by Kronik et al. (1998) [1], which explains the enhanced photovoltaic efficiency of sodium containing CuInSe2 solar cells by the passivation of selenium vacancies at grain boundaries. & 2011 Elsevier B.V. All rights reserved.
Keywords: Thin-film solar cells Photovoltaic Cu(In,Ga)Se2 CuInSe2 Sodium diffusion Grain boundary segregation Pulsed laser atom probe Transmission electron microscopy
1. Introduction Copper indium diselenide (CIS), a compound semiconductor belonging to the I–III–VI2 chalcopyrite family, is one of the most promising absorber materials for the fabrication of thin-film solar cells. CIS possesses several beneficial material properties such as a direct band gap (Eg ¼1.02 eV at room temperature), high absorption coefficient ( Z105 cm 1), long-term stability, and high radiation hardness [2]. CIS solar cells are commonly deposited on soda lime glass (SLG) substrates, which comprise SiO2, Na2O, K2O, and CaO. Several authors reported diffusion of alkali impurities, in particular sodium, into the CIS film during the processing of these solar cells [3,4]. Interestingly, sodium impurities were found to significantly enhance the photovoltaic conversion efficiency (mainly the open circuit voltage) of CIS solar cells, also known as the ‘‘sodium effect’’ [5,6]. An increase in p-type conductivity [7] and a more pronounced texture (predominance of uniaxial crystallite orientation) [8] were suggested as possible reasons for the enhanced efficiency due to sodium incorporation. However, the ‘‘sodium effect’’ has remained a subject of controversial discussion. To develop a comprehensive theory of how sodium improves the conversion efficiency of CIS solar cells, one needs to study the elemental distribution at the atomic scale. The concentrations in bulk and grain boundaries are of particular interest. Some attempts have been previously made to determine the impurity concentrations in CIS absorber layers. For instance, Niles et al. [9] applied Auger Electron Spectroscopy (AES) to detect sodium and oxygen segregation at the surface and grain boundaries of CIS and
n
Corresponding author. E-mail address: [email protected] (O. Cojocaru-Mire´din).
0304-3991/$ - see front matter & 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.ultramic.2010.12.034
to show that solute concentrations inside the grains are highly dilute. However, the detection limit of AES for both sodium and oxygen was reported to be only about 0.15 at% [9]. Among the nano-analytical techniques, Atom Probe Tomography (APT) [10–13] is one of the most promising tools for studying solar cells, as it allows for spatially resolved chemical analyses with subnanometer resolution [14] and the detection of dilute impurity concentrations as low as a few tens of ppm [15]. First APT results on the elemental distribution in Cu(In,Ga)Se2 layers, which are similar to CIS, were recently published by Cadel et al. [15] and Schlesiger et al. [16]. Both authors reported significant sodium enrichment at the grain boundaries. The current paper elucidates the atomic-scale elemental distribution in an actual CIS solar cell, using pulsed laser APT. The enhanced energy conversion efficiency due to sodium incorporation is discussed on the basis of the obtained APT results.
2. Experimental ¨ SonnenA CIS solar cell was fabricated at the Zentrum fur energie- und Wasserstoff-Forschung in Stuttgart, Germany. Soda lime glass (SLG) of 3 mm in thickness was used as a substrate material. A Mo back contact layer of 500 nm was deposited onto the SLG substrate by DC sputtering. Subsequently, a polycrystalline CuInSe2 film of 2 mm in thickness was grown by co-evaporation of the constituent elements using a high temperature single-stage inline process [17]. The substrate temperature during film growth was kept at 600 1C for about 30 min. A CdS buffer layer was deposited by means of chemical bath deposition at 65 1C followed by RF sputtering of an intrinsic ZnO layer, DC sputtering of a ZnO:Al front contact layer, and electron beam evaporation of Ni/Al contact grids.
O. Cojocaru-Mire´din et al. / Ultramicroscopy 111 (2011) 552–556
X-ray fluorescence spectrometry (XRF) measurements were performed to determine the overall composition of the CIS layer, using an EAGLE XXL system (0.1 mbar, Si(Li) detector and 50 kV Rh X-ray source). Using the Cu, In, and Se K lines, the accuracy of the calculated concentrations is better than 0.5 at%. Structural and compositional analyses of CIS films were done by scanning transmission electron microscopy (STEM) and atom probe tomography (APT), respectively. For the STEM investigations, cross-sectional samples were prepared using a dual-beam focused-ion-beam (FIB) (FEI Helios Nanolab 600) and the lift-out technique [18]. STEM was done at a JEOL 2200 FS microscope operated at 200 kV. APT samples were also prepared by means of FIB according to the procedure described in [19,20]. In order to minimize beam damage, a low energy (5 keV) Ga beam was used at the final ion-milling stage. All APT experiments were carried out applying laser pulses of 532 nm wavelength, 12 ps pulse length, and an energy of 0.1 nJ per pulse at a repetition rate of 100 kHz. The specimen base temperature was about 60 K.
3. Results and discussion 3.1. Microstructure of the CIS layer Fig. 1 shows a cross-sectional STEM bright-field image of the studied CIS solar cell, where the stacking sequence of the constituent layers is ZnO:Al/ZnO/CdS/CuInSe2/Mo/SLG. The STEM image reveals
Fig. 1. STEM bright-field image of a CuInSe2-based thin-film solar cell in a crosssectional view.
553
a polycrystalline CIS layer, which is not smooth but exhibits substantial roughness. Top-view scanning electron microscopy images (not shown here) revealed faceted grains of large size (between 0.7 and 1.25 mm) and are in agreement with STEM observations. Besides the large grains in the top CIS region, small grains of 0.13 to 0.36 mm in size can be recognized at the bottom of the CIS layer (see Fig. 1). Such a bimodal grain structure is typical of CIS films grown by elemental co-evaporation in the presence of sodium [21] and can be discussed in terms of the van der Drift evolutionary selection mechanism [22]. In this mechanism, the nucleation stage is characterized by randomly oriented nuclei that initially grow independently at uniform rate. In the following stage, the crystals with the direction of fastest growth perpendicular to the substrate are favored over other crystals and will prevail. For a tetragonal crystal, the directions of fastest growth are /0 0 1S, /1 0 0S, and /1 1 0S [22]. Another interesting feature is the presence of a few voids (about 10 to 100 nm in size) in the CIS layer close to the CIS/Mo interface (see Fig. 1). These voids probably stem from the deposition process and are not formed during TEM sample preparation, as the sample was finally ion-milled at only 5 kV. Such nano-voids in the CIS layer were also observed by Lei et al. [23] and for CIGS layers by Rudmann [24].
3.2. Composition of CIS layer (grain and grain boundary) A typical mass spectrum obtained from an APT analysis of a CIS layer is given in Fig. 2. While copper and indium were detected as both single and double charged ions, selenium ions were only detected in a single charged state. Complex ions, namely CuCO + , CuSe + , Cu2Se + , CuSe2+ , Se2+ , and Se3+ were detected as well. Sodium was detected as Na + . The background noise level in the mass range close to the Na + peak is extremely low (E40 ppm/a.m.u.) so that even very low dilute sodium concentrations can be detected. Besides sodium, oxygen and silicon impurities were detected as O + , O2+ , Si2 + , and Si + . A small amount of Ga + was also detected. Fig. 3 shows three-dimensional maps of all the detected elements as well as sodium and oxygen only. The analyzed volume is about 120 120 390 nm3 in size and represents the interior of a CIS grain. The measured bulk composition is 23.470.06 at% Cu, 29.070.07 at% In, and 47.470.08 at% Se, which is in good agreement with the integral XRF measurement (compare APT-1 and
Fig. 2. Mass spectrum of the complete measurement from Fig. 3 with details showing the minority elements (Na and O) in the low mass region.
O. Cojocaru-Mire´din et al. / Ultramicroscopy 111 (2011) 552–556
554
XRF in Table 1). In the following, APT-1 and APT-2 denotes the APT measurements presented in this work. While the sodium concentration detected by APT for the CIS bulk is only about 0.002 at% (20 ppm), the oxygen concentration is substantially higher and is about 0.07 at% (700 ppm). The sodium concentration measured in this work is significantly lower than the value reported by Cadel et al. [15] for a Cu(In,Ga)Se2 (CIGS) film (85 ppm). Furthermore, Cadel et al. [15] did not report any oxygen inside CIGS. The difference in the results between these two works may well be ascribed to different processing conditions as well as to different compositions of the absorber layers (CIS and CIGS). In Fig. 3 neither sodium nor oxygen clusters can be distinguished. If such clusters exist they are so tiny that they cannot be clearly recognized in the elemental maps. Thus, to verify or refute a possible clustering trend or short range ordering of sodium and/ or oxygen atoms, the cluster identification (maximum separation method [25]) and the first nearest neighbor (abbreviated as 1NN) distance distribution method were applied. Neither sodium nor oxygen clusters were identified by the cluster identification, even by systematically varying Nmin (minimum number of atoms per cluster) and dmax (maximum distance between sodium and/or oxygen atoms) values. The measured 1NN distance distribution for the oxygen atoms, which is compared to a randomized oxygen distribution in Fig. 4, should therefore be interpreted as a random oxygen distribution. The same trend is observed for Na (not shown here). Fig. 5a shows another APT data set of the CIS layer. In these elemental maps, we clearly recognize sodium, oxygen, and silicon enriched zones, and also a strong correlation between the distributions of these elements. Rotation and careful inspection of these enrichment zones reveal that they have a two-
Fig. 3. APT elemental maps of copper (blue), indium (pink), selenium (red), gallium (yellow), sodium (maroon) and oxygen (sky blue). The volume size shown in this figure is 120 120 390 nm3. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
dimensional shape. However, these enrichments are not continuous but are divided into four plate-shaped parts, marked in Fig. 5a by a dashed-line (for more details see Ref. [26]). Previous AES studies revealed a correlation between sodium and oxygen enrichments (in agreement with Fig. 5a) at the CIS grain boundaries [9]. Taking these previous results and the two-dimensional shape of the enrichments into account, we may conclude that the solute enrichments seen in Fig. 5a are located at or close to a grain boundary. The measured bulk composition (excluding the grain boundary zone) is 21.070.06 at% Cu, 28.8 70.07 at% In, and 50.0 70.08 at% Se, which is in better agreement with the integral XRF measurement (compare APT-2 and XRF in Table 1) than the APT-1 composition. Our APT measurements indicate compositional fluctuations within the CIS film. This result is in agreement with studies done by Eich et al. [29] on lateral inhomogeneities of a Cu(In,Ga)Se2 absorber film. Slight variations in sodium and oxygen concentrations in the bulk compared to APT-1 can be measured as well (2675 ppm of sodium and 600 710 ppm of oxygen, respectively). Although sodium and oxygen atoms are strongly enriched at the grain boundaries, their concentration rapidly drops to the bulk values, when moving away from the segregation zone. In the bulk, sodium and oxygen atoms are homogeneously distributed over several hundreds of nanometers, indicating that long range diffusion has taken place. These observations suggest that the measured bulk concentration values of sodium and oxygen correspond to their solubility limit in CIS. In all APT measurements small amounts of gallium were detected ( 0.1 at%). Gallium atoms stem from the deposition process, as confirmed by secondary ion mass spectrometry (SIMS) ( 0.04 at%). A possible explanation for the relatively high gallium content measured by APT is the implantation of gallium
Fig. 4. Distribution of the 1NN distances for the oxygen atoms, represented in Fig. 3. Comparison with the randomised oxygen distribution. The bin size used for this distribution is 0.2 nm.
Table 1 Chemical composition of single CIS grains measured by APT and of the whole CIS layer (integral) measured by XRF; APT-1: composition of grain shown in Fig. 3 and APT-2: composition of grain shown in Fig. 5 without GB area. C(at%)
CCu
CIn
CSe
CNa
CO
CGa
CSi
APT-1 APT-2 XRF
23.4 7 0.06 21.0 7 0.06 21.3
29.07 0.07 28.8 70.07 27.7
47.4 7 0.08 50.0 70.08 50.8
0.0027 0.0005 0.00267 0.0005 –
0.077 0.001 0.067 0.001 –
0.12 0.11 0.02
– – –
O. Cojocaru-Mire´din et al. / Ultramicroscopy 111 (2011) 552–556
555
of a grain boundary ( 0.5 nm), which can be partly ascribed to a local magnification effect [27]. This reconstruction artefact results in a magnified width and lower atomic density of the oxygen rich grain boundary region, which has a higher evaporation field than the CIS matrix. Letellier [28] observed for nickel-based superalloys that the influence of the local magnification effect increases continuously as the grain boundary approaches an orientation parallel to the sample tip axis. Thus, the enlarged width of the oxygen rich grain boundary zone in Fig. 5a and b is not surprising. The Gibbsian interfacial excess of solutes G (number of segregated atoms per unit interfacial area) was determined from the proxigrams following the procedure described in [29,31]. We obtain GO ¼4.7 1019 atoms/m2, GNa ¼1.4 1018 atoms/m2, and GSi ¼2.8 1018 atoms/m2. These values correspond to nearly 10, 0.25, and 0.5 monolayers of oxygen, sodium, and silicon, respectively, with respect to the (1 0 0) CIS plane, where 10 monolayers of oxygen correspond to 5 nm. Hence, an oxygen containing compound phase may have been formed at the grain boundary. It is not possible to determine the exact composition of the oxygen rich zones because of the strong local magnification effect. However, a strong depletion of selenium can be clearly detected in conjunction with the oxygen enrichment, strongly suggesting that a large number of vacant selenium sites at the grain boundaries are occupied by oxygen atoms. It should be mentioned that similar trends could be measured for several grain boundaries using EDX (not shown here). APT analyses of other grain boundaries confirmed sodium and oxygen segregation. On the other hand, compositional differences at the grain boundaries could be detected as well. For instance, the slight copper enrichment and indium depletion seen in Fig. 5b is not generally observed. The difference in the measured compositions of various grain boundaries is not surprising in view of the variation in solute contents as a function of grain boundary misorientation [32].
3.3. Correlation between elemental distribution and enhanced photovoltaic efficiency
Fig. 5. (a) APT elemental maps of copper (blue), indium (pink), selenium (red), gallium (yellow), sodium (maroon), oxygen (sky blue) and silicon (black). The volume size shown is 72 72 260 nm3. (b) Proxigram of species in the sample with respect to the O 7% isoconcentration surface; the bin size is 0.2 nm. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
during focused-ion-beam milling. However, we note that the implantation zone in CIS specimens is restricted to a few nanometers below the surface if 5 keV ions are used in the final ionmilling step. Fig. 5b shows a proxigram (for more details see Ref. [30]) of the elements detected in the sample with respect to the 7 at% oxygen isoconcentration surface. The proxygram clearly reveals sodium, oxygen, and silicon accumulations at a CIS grain boundary, where the oxygen concentration is substantially higher ( 18 at%) than the sodium ( 0.6 at%) and silicon (1.4 at%) concentration ( 1 at%). These solute enrichments are most likely due to out-diffusion from the soda lime glass substrate during the deposition process (at 600 1C for 30 min), as they were not found in APT measurements of CIGS layers grown on steel substrates (not shown here, to be published). The width of the detected oxygen enrichment is about 7 to 8 nm and therefore much larger than the typical structural width
Our APT data clearly reveal that sodium, silicon, and oxygen atoms diffuse from the SLG substrate into the CIS film during film growth. The enrichment of sodium and oxygen at the grain boundary indicates that diffusion into the CIS film occurs mainly along the grain boundaries, where defect density is higher than in the bulk. Highly dilute concentrations of sodium ( 20 ppm) and oxygen ( 700 ppm) are detected inside the CIS grains. These solutes could passivate detrimental lattice defects (e.g., donor like defects like selenium vacancies, indium atoms at copper sites, etc.) and contribute to the enhancement of the solar cell efficiency. However, the dilute concentrations of the solutes in the grains indicate the ‘‘sodium effect’’ cannot be purely regarded as a bulk effect. As presented in Fig. 5 grain boundary enrichments of oxygen and sodium are accompanied by a strong depletion of Se. These observations are in agreement with the model proposed by Kronik et al. [1] explaining the ‘‘sodium effect’’ on the basis of the well-known catalytic effect of alkali metals on surface oxidation of semiconductors [33]. O2 can be formed from physisorbed O2 by alkali-metal induced polarization of the O–O bond and lowering of the surface work function. According to Ref. [1], the formation of grain boundaries in CIS gives rise to a large number of Se vacancies, which introduce intragap donor levels. These donors lower the effective p-type conductivity. Neutralizing the selenium vacancies by chemisorbed O2 according to the reaction + [In0–V+Se ] +O2 -[In0–O0Se], the photovoltaic efficiency can be increased. Once formed, the In–O bonds are stable and can only
556
O. Cojocaru-Mire´din et al. / Ultramicroscopy 111 (2011) 552–556
be broken by significant and deliberate reduction. Furthermore, this defect chemical model and our APT results suggest that not only grain boundary passivation is enhanced by the presence and catalytic effects of sodium but also sodium diffusion can be assisted by oxygen.
4. Conclusions In this work, Atom Probe Tomography was employed to investigate the elemental distribution in the CuInSe2 absorber of a thin-film solar cell. Sodium and oxygen atoms were found to be distributed homogeneously in the CuInSe2 grains. The measured concentrations of sodium ( 0.002 at%) and oxygen ( 0.07 at%) may well correspond to the solubility limit of sodium and oxygen in CuInSe2 at 600 1C. Furthermore, a correlated enrichment of sodium, oxygen, and silicon atoms was detected, presumably at a grain boundary. These enrichments are accompanied by selenium depletion, which is in good agreement with the defect chemical model proposed by Kronik et al. [1]. The ‘‘sodium effect’’ in our CIS-based solar cell could be explained by the passivation of selenium vacancies (detrimental donor defects) by oxygen.
Acknowledgements The authors thank Wolfram Witte for the CIS-based solar cell preparation and also Aleksander Kostka and Axel Eicke for their help in this work. References [1] L. Kronik, D. Cahen, H.W. Schock, Adv. Mater. 10 (1998) 31. [2] Hegedus, Antonio Luque, Steven, Handbook of Photovoltaic Science and Engineering, England: s.n., 2003. ¨ ˚ [3] J. Hedstrom, H. Ohlsen, M. Bodegard, A. Kylner, L. Stolt, D. Hariskos, M. Ruckh, H.W. Schock, Conf. Rec. 23rd IEEE Photovolt. Spec. Conf. IEEE, Piscataway, N.J., 1993, p. 364. ¨ [4] L. Stolt, J. Hedstrom, J. Kessler, M. Ruckh, K.-O. Velthaus, H.-W. Schock, Appl. Phys. Lett. 62 (1993) 597. ˚ [5] K. Granath, M. Bodegard, L. Stolt, Sol. Energy Mater. Sol. Cells 60 (2000) 279–293. ˚ [6] M. Bodegard, K. Granath, L. Stolt, Thin Solid Films 361 (2000) 9–16.
[7] J. Holz, F. Karg, H. von Philipsborn, in: Proceedings of 12th European Photovoltaic Solar Energy Conference, Amsterdam: s.n., 1994, p. 1592. [8] B.J. Stanbery, Critical Rev. Sol. Mater. Sci. 27 (2) (2002) 73–117. [9] D.W. Niles, M. Al-Jassim, K. Ramanathan, J. Vac. Sci. Technol. A 17 (1) (1999) 291. [10] A. Cerezo, T.J. Godfrey, S.J. Sijbrandij, G.D.W. Smith, P.J. Warren, Rev. Sci. Instrum. 69 (1998) 49. [11] M.K. Miller, T.F. Kelly, Rev. Sci. Instrum. 78 (2007) 031101. [12] B. Deconihout, F. Vurpillot, B. Gault, G. Da Costa, M. Bouet, A. Bostel, A. Hideur, G. Martel, M. Brunel, D. Blavette, Surf. Interf. Anal. 39 (2007) 278–282. [13] B. Gault, F. Vurpillot, M. Gilbert, A. Vella, A. Menand, D. Blavette, B. Deconihout, Rev. Sci. Instrum. 77 (2006) 043705. [14] B. Gault, M.P. Moody, F. De Geuser, A. La Fontaine, L.T. Stephenson, D. Haley, S.P. Ringer, Microsc. Microanal. 16 (2010) 99–110. [15] E. Cadel, N. Barreau, J. Kessler, P. Pareige, Acta Mater. 58 (2010) 2634–2637. ¨ [16] R. Schlesiger, C. Oberdorfer, R. Wurz, G. Greiwe, P. Stender, M. Artmeier, P. Pelka, F. Spaleck, G. Schmitz, Rev. Sci. Instrum. 81 (2010) 043703. [17] M. Powalla, G. Voorwinden, B. Dimmler. in: Proceedings of the 14th European Photovoltaic Solar Energy Conference, Barcelona, Spain: s.n., 1997, p. 1270. [18] L.A. Giannuzzi, J.L. Drown, S.R. Brown, R.B. Irwin, F.A. Stevie, Microsc. Res. Tech. 41 (4) (1998) 285–290. [19] M.K. Miller, K.F. Russell, K. Thompson, R. Alvis, D.J. Larson, Microsc. Microanal. 13 (2007) 428–436. [20] K. Thompson, D. Lawrence, D.J. Larson, J.D. Olson, T.F. Kelly, B. Gorman, Ultramicroscopy 107 (2007) 131–139. ˚ [21] M. Bodegard, O. Lundberg, J. Lu, L. Stolt, Thin Solid Films 431–432 (2003) 46–52. [22] A. van der Drift, Philips Res. Rep. 22 (1967) 267–288. [23] C. Lei, A. Rockett, I.M. Robertson, J. Appl. Phys. 100 (2006) 073518. [24] D. Rudmann, Effects of sodium on growth and properties of Cu(In, Ga)Se2 ¨ thin films and solar cells, Ph.D Thesis, Eidgenossische Technische Hochschule ¨ ETH Zurich, Nr. 15576, 2004. [25] M.K. Miller, Atom Probe Tomography, [Ed.] Academic/Plenum Press, New York: s.n., 2000. [26] O. Cojocaru-Miredin, P. Choi, R. Wuerz, D. Raabe, in: S. Vlase, and A. Chiru D. Bigoni, [Eds.], Proceedings of Third International Conference on Advanced Composite Materials in Civil Engineering, Brasov/Romania: Transilvania University, 2010,’’ vol. 1, p. 45. [27] F. Vurpillot, A. Cerezo, D. Blavette, D.J. Larson, Microsc. Microanal. 10 (3) (2004) 384–390. [28] L. Letellier, Etude des joints de grains et interphases dans les superalliages Astroloy par microscopie electronique et tomographie atomique, University of Rouen, Rouen, France, 1994. [29] D. Eich, U. Herber, U. Groh, U. Stahl, C. Heske, M. Marsi, M. Kiskinova, W. Riedl, R. Fink, E. Umbach, Thin Solid Films 361–362 (2000) 258–262. [30] O.C. Hellman, J.A. Vandenbroucke, J. Rusing, D. Isheim, D.N. Seidman, Microsc. Microanal. 6 (2000) 437–444. [31] O.C. Hellman, D.N. Seidman, Mater. Sci. Eng. A 237 (2002) 24–28. [32] A. Biscondi, M. Fraczkiewicz, J. Physique 46 (1985) 497. [33] H.I. Starnberg, P. Soukisasian, Z. Hurych, Phys. Rev. B 39 (1989) 12775.